[Bug] trilu not tagged with injective and thus miss reduce schedule

[ganler]

Expected behavior

TVM should successfully compile a model whose operators are supoprted.

Actual behavior

For some cases, trilu-argmin cannot find a reduction strategy and thus throw:

tvm/python/tvm/topi/x86/reduction.py

Line 90 in 8131364

raise RuntimeError("Unsupported operator: %s" % operator.tag)

The reason is that in this case trilu's operator.tag is empty which is not regarded as injective:

tvm/python/tvm/topi/x86/reduction.py

Line 84 in 8131364

if tag.is_injective(operator.tag):

I am not sure but maybe we need to put proper operator tag for trilu?

Environment

fa17da2 on Ubuntu 20.04.

Steps to reproduce

Preferably a minimal script to cause the issue to occur.

import tvm
from tvm import relay

i0 = relay.var("i0", shape=[2, 2], dtype="int32")
i1 = relay.var("i1", shape=(), dtype="int64")
v0 = relay.trilu(i0, i1)
v1 = relay.argmin(v0, axis=[0])

f = relay.Function([i0, i1], v1)
relay.create_executor("graph", device=tvm.cpu(), target="llvm").evaluate(f)

Triage

Please refer to the list of label tags linked above to find the relevant tags and add them here in a bullet format (example below).

• needs-triage

cc: @jwfromm